Post

The next AI trade may not be about memory anymore. It may be about light. 👀 Wall Street is starting to ask a very uncomfortable question: What if the biggest AI bottleneck is no longer storage—but the connection between GPUs? For the past year, the market has been obsessed with HBM and DRAM shortages. SK Hynix was one of the biggest beneficiaries of that narrative. But the story may be changing. A controversial take from Citrini analyst @jukan05 has already crossed 2 million views: “Short storage, long optics.” His argument is pretty simple—and honestly, worth thinking about. First, forced selling in South Korean leveraged ETFs could create significant passive selling pressure on storage names. Second, and more importantly, Nvidia may increasingly rely on optical interconnects to connect multiple racks and build massive AI clusters, rather than simply pushing more HBM onto every individual GPU. And third, the market is increasingly expecting memory prices to peak within the next couple of quarters. If that happens, the AI hardware profit pool could start moving. Think about it this way: HBM helps the GPU think faster. Optics helps thousands of GPUs talk to each other fast. That could mean the certainty behind the memory trade starts fading, while the optical-interconnect story becomes much stronger. And this is exactly why the market is beginning to watch optics so closely. 🇨🇳 China: $300308 Zhongji Xuchuang and $300502 Xinyisheng are two of the biggest names in optical modules, while $300394 Tianfu Communication sits further upstream in precision optical components. 🇺🇸 US: COHR and LITE are important players in optical chips and lasers, while AAOI has become one of the more aggressive optical-module plays—with much higher volatility. 🇰🇷 South Korea: Storage remains the country's strongest AI-hardware story, but its optical ecosystem doesn't have an obvious global heavyweight comparable to the leaders in China and the US. #DailyOrbit